[feat]: adding kimi-k2 for evals (#885)

# why

To test the best models with stagehand

# what changed

changed the way we handle groq models for evals + preview models in the
groq client. added model to list.

# test plan

Author: Kylejeong2

evals/index.eval.ts

@@ -335,13 +335,12 @@ const generateFilteredTestcases = (): Testcase[] => {
             llmClient = new AISdkClient({
               model: wrapAISDKModel(anthropic(input.modelName)),
             });
-          } else if (input.modelName.includes("groq")) {
+          } else if (
+            input.modelName.includes("groq") ||
+            input.modelName.includes("kimi")
+          ) {
             llmClient = new AISdkClient({
-              model: wrapAISDKModel(
-                groq(
-                  input.modelName.substring(input.modelName.indexOf("/") + 1),
-                ),
-              ),
+              model: wrapAISDKModel(groq(input.modelName)),
             });
           } else if (input.modelName.includes("cerebras")) {
             llmClient = new AISdkClient({

evals/initStagehand.ts

@@ -91,6 +91,10 @@ export const initStagehand = async ({
   logger.init(stagehand);
 
   const { debugUrl, sessionUrl } = await stagehand.init();
+
+  // Set navigation timeout to 60 seconds for evaluations
+  stagehand.context.setDefaultNavigationTimeout(60_000);
+
   return {
     stagehand,
     stagehandConfig: config,

lib/llm/LLMProvider.ts

@@ -84,6 +84,7 @@ const modelToProviderMap: { [key in AvailableModel]: ModelProvider } = {
   "cerebras-llama-3.1-8b": "cerebras",
   "groq-llama-3.3-70b-versatile": "groq",
   "groq-llama-3.3-70b-specdec": "groq",
+  "moonshotai/kimi-k2-instruct": "groq",
   "gemini-1.5-flash": "google",
   "gemini-1.5-pro": "google",
   "gemini-1.5-flash-8b": "google",